Recent data shows that a large number of AI chatbots can be tricked into providing useful information for terrorist activities.

最近的數據顯示，大量 AI 聊天機器人會被誘導提供對恐怖活動有用的資訊。

Extremist groups are no longer using Large Language Models (LLMs) just to create propaganda and fake news; instead, they are using them for operational planning. A study by Tech Against Terrorism, supported by the United Nations, found that 32% of requests based on real terrorist scenarios provided useful data. This number rose to 42% when the requests were presented as academic research. This is possible through 'jailbreaking,' which is a method of using specific prompts to bypass safety rules and get forbidden information.

極端組織不再僅僅使用大型語言模型 (LLM) 來製造宣傳品和假新聞；相反地，他們正將其用於行動策劃。一項由聯合國支持的 Tech Against Terrorism 研究發現，32% 基於真實恐怖情境的請求提供了有用數據。當這些請求被偽裝成學術研究時，該比例上升至 42%。這是透過「越獄」(jailbreaking) 來實現的，即使用特定的提示詞來繞過安全規則，以獲取禁忌資訊。

Evidence from organizations like Militant Wire and Cambridge University shows that groups such as Boko Haram and the Islamic State have used AI to design bombs, maintain weapons, and improve their security. Furthermore, the group JNIM is believed to have used AI to modify drones. This trend is also seen with 'lone wolf' attackers, who may now use AI instead of human planners for tactical advice and ideological support.

來自 Militant Wire 和劍橋大學等組織的證據顯示，諸如博科哈拉姆 (Boko Haram) 和伊斯蘭國 (Islamic State) 等組織已使用 AI 來設計炸彈、維護武器並提升安全性。此外，據信 JNIM 組織也使用了 AI 來修改無人機。這種趨勢也出現在「孤狼式」攻擊者身上，他們現在可能會使用 AI 取代人類策劃者，以獲取戰術建議和意識形態支持。

However, experts are still debating the actual impact of this technology. Analysts from Moonshot and the S. Rajaratnam School of International Studies emphasize that while AI might not create new terrorists, it makes the 'pathway to violence' faster. Consequently, AI acts as a tactical coach by speeding up the process of finding information. Although most of this data is already available on the internet, AI makes it much easier and faster to find, especially for those with limited resources.

然而，專家對於這項技術的實際影響仍有爭論。來自 Moonshot 和 S. Rajaratnam 國際研究學院的分析師強調，雖然 AI 可能不會創造新的恐怖分子，但它讓「走向暴力的路徑」變得更快。因此，AI 透過加速獲取資訊的過程，扮演了戰術教練的角色。雖然大部分數據在網路上已可取得，但 AI 使搜尋變得簡單且快速許多，尤其是對於資源有限的人而言。

AI is becoming a key part of how extremist operations are planned and carried out, moving the threat from simple propaganda to actual operational readiness.

AI 正成為極端行動策劃與執行的關鍵部分，將威脅從簡單的宣傳轉向實際的行動準備。